[PATCH v2 2/4] phy: qcom: qmp-pcie: Add qref and refgen regulator vote for QCS8300 and SA8775p PHY

[Konrad Dybcio]

On 9/16/26 12:20 PM, Ziyue Zhang wrote:
> Add a new sa8775p_qmp_phy_vreg_l that includes vdda-phy, vdda-pll,
> vdda-qref, vdda-refgen and refgen supplies, and use it for QCS8300
> and SA8775p PCIe PHY configurations. This avoids modifying
> sm8550_qmp_phy_vreg_l and breaking SM8550 and SM8650.
> 
> These supplies were missing from the original driver, which was found
> to be the root cause of occasional system crashes observed on QCS8300
> and SA8775p platforms. Without voting for these supplies, the PCIe PHY
> lacks stable reference clock and voltage, so the PHY never becomes
> ready and initialization times out:
> 
>   [    7.303119] qcom-qmp-pcie-phy 1c04000.phy: phy initialization timed-out
>   [    7.309948] phy phy-1c04000.phy.0: phy poweron failed --> -110
> 
> Note that due to a hardware issue, QREF actually depends on refgen3
> rather than refgen2 as documented; refgen3 is therefore voted manually
> via the refgen supply as a workaround.
> 
> Fixes: a05b6d5135ec ("phy: qcom-qmp-pcie: add support for sa8775p")
> Fixes: ebf198f17b5a ("phy: qcom-qmp-pcie: add dual lane PHY support for QCS8300")
> Signed-off-by: Ziyue Zhang <ziyue.zhang at oss.qualcomm.com>
> ---
>  drivers/phy/qualcomm/phy-qcom-qmp-pcie.c | 16 ++++++++++------
>  1 file changed, 10 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/phy/qualcomm/phy-qcom-qmp-pcie.c b/drivers/phy/qualcomm/phy-qcom-qmp-pcie.c
> index da83358c97ea..ab81c6e584a8 100644
> --- a/drivers/phy/qualcomm/phy-qcom-qmp-pcie.c
> +++ b/drivers/phy/qualcomm/phy-qcom-qmp-pcie.c
> @@ -4094,6 +4094,10 @@ static const char * const sm8550_qmp_phy_vreg_l[] = {
>  	"vdda-phy", "vdda-pll", "vdda-qref",
>  };
>  
> +static const char * const sa8775p_qmp_phy_vreg_l[] = {
> +	"vdda-phy", "vdda-pll", "vdda-qref", "vdda-refgen", "refgen",
> +};

I think we could grab them unconditionally, as QREF/REFGEN appears
on most/all SoCs (perhaps with regulator_optional_get so that we
defer the correctness of this to dt-bindings for any specific
platform)

Konrad